An Architecture for Vehicle Diagnostics in Software-Defined Vehicles

Vehicles are evolving into Software-Defined Vehicles. The increasing use of automotive High Performance Computers (HPCs) provides more computing power and storage resources in vehicles. This opens possibilities to use more in-vehicle software. However, it also leads to challenges for vehicle diagnostics. Today's diagnostic approaches, based on Diagnostic Trouble Codes (DTCs), are not suitable for software on HPCs. For example, this software is highly variable and updated over time, so predefined DTCs are not dynamic enough. This introduces a degree of ambiguity into the diagnostic processes. Additional diagnostic data are required. In the Cloud, observability approaches are becoming widely used for software. Observability involves examining the availability and performance of an entire software system. To detect failures early, observability data, such as logs, metrics, and traces, are used. This is of interest for vehicle diagnostics as new diagnostic approaches are needed to continuously monitor and observe software in vehicles. Observability data could be used to detect deviations and failures in software systems, in ECUs, and the whole vehicle at an early stage as well as to identify the failure causes. Therefore, this paper describes a vehicle diagnostic architecture for providing observability data in vehicles. The focus of the data provisioning relies on the software on the HPCs. The architecture is evaluated with two vehicle diagnostic expert surveys and a set of expert interviews. In addition, the usage of observability data for vehicle diagnostics is discussed based on the results from the different vehicle diagnostic experts as well as how it enables future vehicle diagnostics of Software-Defined Vehicles.